v2.0 Release

This release mainly updates the RSTFinder code for compatibility with the latest versions of SKLL (and, by extension, scikit-learn).

• Patch RST parser model pruning to apply correctly to newer scikit-learn estimators (#72).
• Update parser model training and inference to use the latest SKLL API (#73).
• Fix a bug where fewer than expected items were returned in an edge case (#74).
• Fix cosmetic issues in code identified by black, flake8, and pydocstyle.

IMPORTANT: Please note the major version change. As newer versions of SKLL are backwards incompatible, so is RSTFinder. You should not use RSTFinder v2.0 on models trained with v1.0.

v1.0 release

This marks the v1.0 release which is a major, new release after a few years of inactivity.

• The package and the repository are no now officially renamed as rstfinder.
• Add detailed to all modules, functions, classes, and methods.
• Update dependencies to latest versions.
• Modernize codebase to work with Python 3.6 – 3.8 and drop Python 2 support.
• Add several new unit and functional tests.
• Add CI integration for automated testing.
• Add a conda recipe and release a conda package for easier installation.
• Update README to include detailed instructions for: (a) installation (b) training new models (c) using trained models.
• Add new functionality that allows users to bring their own (properly-formatted) constituency trees as the sole input (see rst_parse.from_constituency_trees()).

0.2.1: Merge pull request #36 from mheilman/develop

added checks on working directories to avoid confusion during parser training